Estimation Of Differences Of Restricted Mean Life Under Dependent Censoring

In clinical medicine and epidemiological studies,the average life expectancy of a patient after a certain treatment is often concerned.Due to the existence of censorship,the tail of the survival function has a bias.When comparing the survival time between two treatment groups,the more restricted mean life may be more suitable,that is,the survival time is limited to the expectation of time instead of life expectancy.In observational studies,when treatment options are not randomly assigned to patients,we also need to consider treatment imbalances among confounding factors.In this paper,we present estimates of difference of the restricted mean life between two different treatment regimens.In view of the difference in treatment efficacy between non-randomized groupings,we consider the problem of inferring the limited mean life difference in the survival time with independent and dependent censoring.The first part of the article uses two models to explain the two types of confounding factors,establishes a proportional hazard model to explain the baseline covariates,and establishes an additive hazard model to explain the time-dependent covariates.The model parameters are estimated and the large sample properties of the estimators are discussed.The second part explains the two types of confounding factors by building an additive-multiplicative risk model,and gives the model parameter estimates and the estimated large sample properties in the same way as in the first part.The performance of the model parameter estimation methods involved in the two parts in terms of skewness and accuracy is given by random simulation.Finally,apply the method we have given to the analysis of the efficacy difference between the two treatments in patients with liver cirrhosis...